Re: policy support [was Re: load-balancing]

[Geoff Huston <gih@telstra.net>]

>The problem is the size of global routing table increased by
>multihoming, regardless of whatever the requirements for
>multihoming are.
>
>The goal is to suppress the size of global routing table.

In which case the first pass corollary would be to declare multi-homing
at any level (host, site, network, region) evil and bad and insist
on adoption of massive levels of hierarchy in connectivity.

I'm not sure that's precisely what we are after - I thought we were after
an approach which would allow multi-homing, but also address some of
the issues we are seeing in the V4 space where multi-homing appears to
be driving a level of growth in the global routing table which may
well be unsustainable. Very quickly this becomes an examination
of routing tools - where the objective is to define routing
tools which operate efficiently across an increasingly dense
semi-meshed policy-rich topologies.

A modest attempt to describe this is at
http://www.telstra.net/gih/papers/ietf/draft-iab-bgparch-00.txt
if you are interested.

thanks,

   Geoff